Where do I see whether the trigger for error spikes fired and which run it woke?

Tares shows whether your error spikes trigger fired on the trigger’s own page and under Firings. Open Activity to see the firing and the run it woke as one thread, or open the agent’s Runs to see every run. A trigger evaluates as events arrive, so if it did not fire, check that its sources and filters match the events.

How it works

A trigger in Tares watches one or more sources and evaluates its condition as events arrive. When the condition holds, the trigger fires and wakes the agent subscribed to it. Each step leaves a record you can open in the console:

• The trigger’s page lists the firings for that trigger. In the demo from the AI SRE guide, the incident trigger’s page is at http://127.0.0.1:8787/triggers/incident.
• Firings is the console list of every firing across triggers.
• Activity shows a firing and the run it woke as one thread. You find it under Setup, then Advanced setup.
• Agents, then your agent, then Runs lists every run of that agent.
• Overview and Explore show the finding the run wrote back. The project’s Overview lists it under What the agents found, and the same finding sits on the service’s timeline in Explore.

The trigger evaluates per service, keyed by the shared service label on your sources. The demo trigger fires when an alert event arrives, using the condition sum(alert_active) > 0 over 1m, per service. Tares does not re-threshold your alerts. Your alerting keeps deciding what is an alert, and Tares ingests what fired and handles the waking.

Set it up

Follow these steps to confirm a firing and find its run. Use your error spikes trigger in place of the demo’s incident trigger.

1. Start Tares and open the console at http://127.0.0.1:8787.

tares up

2. Open the trigger’s page for error spikes, or open Firings. If you see a firing with a time that matches the spike, the trigger fired.
3. Open Setup, then Advanced setup, then Activity. Find the firing in the thread. The run it woke is shown in the same thread.
4. Open Agents, pick the agent subscribed to the trigger, and open Runs. Match the run to the firing by time.
5. Open Overview and look under What the agents found, or open the service in Explore. The finding the run wrote back appears there.
6. If you run the bundled demo and want to see the whole chain, cause an error spike on purpose. The call below is the same one the Cause an incident button makes.

curl -s -XPOST localhost:8080/demo/inject -H 'content-type: application/json' \
  -d '{"scenario": "error_spike"}'

In the demo, the alert fires within about 30 seconds. Then the trigger fires and the agent wakes. Check the trigger’s page, Activity and the agent’s Runs in that order.

If there is no firing, work through these checks:

1. Open Explore for the service and confirm that events from the sources are arriving. The trigger evaluates only as events arrive.
2. Open the trigger and confirm its sources list includes the sources that carry the error events.
3. Check its filters against the events. A filter on a field or value that the events do not carry never matches.
4. Check that the sources are keyed by the same service label, so the events land on one timeline per service.
5. Check the condition and its window against what the events show. The trigger fires only when the condition holds within that window.

What you get

• A clear yes or no on whether the error spikes trigger fired, with the time of each firing.
• The run the firing woke, shown next to the firing in Activity, so you do not have to match timestamps by hand.
• The full list of the agent’s runs under Runs, for the cases where one trigger woke several.
• The finding the run wrote back, on Overview and on the service’s timeline in Explore, so the next investigation starts from that conclusion.
• A short checklist of sources, filters and condition when nothing fired, which tells you where the chain broke.

When something else fits

• If you want to look at the evidence itself rather than the firing, use Explore, or ask a connected MCP client. It reads the same correlated timeline and the findings on it.
• If you are still building the trigger and have not run the setup, the AI SRE guide walks through sources, triggers and agents end to end.

Related